dave kasten (@David_Kasten) profile picture

AI security hawk. "Do what seems cool next." Formerly: McKinsey, VaccinateCA, Activision Blizzard.

Joined March 2011
Tell your agent to use referral code “RSI” for 100% off
1
5
344
In case you're tracking where we are in the superintelligence-risk-as-COVID-2020 metaphor, we're at "Senator retweets Zvi" stage of the operation
Nvidia CEO Jensen Huang Had an Affair With a Chinese Spy, by @TheZvi thezvi.substack.com/p/nvidia…
2
16
231
6,371
Oh.
🚨NEW: Dozens of accounts defending Jensen Huang after his affair with an alleged Chinese spy appear to be fake. They are part of a much wider pro-Nvidia campaign, linked to an Israeli influence firm that says it shut down last year… But the accounts never stopped posting. 🧵
1
22
1,042
Personally, I feel so much more safe when I see the extremely hung over National Guardsmen slumped over wearing their sunglasses inside my very dark Metro station in the darkest possible corner they can find
3
1
1
39
1,166
The current situation: “So basically they invite their friend over to take a look at what they're doing and then their friend says "wow" and tells the rest of us "that was wild, can't tell you the details but there's some really wild stuff going on in there."”
Definitely *not* regulatory markets as described by Jack Clark in his seminal paper on the topic. Merely delegating the most important regulatory task of all time to private companies that are paid and selected by the regulated entities themselves would be unacceptably restrictive. Instead, they want a much more libertarian thing where the private regulator they pay and select is not licensed by the government or accountable to the government in any way, and also has no ability to tell Anthropic to do anything Anthropic doesn't want to do. So basically they invite their friend over to take a look at what they're doing and then their friend says "wow" and tells the rest of us "that was wild, can't tell you the details but there's some really wild stuff going on in there."
1
6
1
72
2,130
This is true but also note that 1. we still need MANY more people and 2. salaries are net moving upwards, despite greater supply; demand effects outweigh
From my vantage point, it feels like a dam has broken and there is a flood of new talent entering the AI safety field! The delta is especially noticeable outside of research roles.
1
17
846
Some key questions: 1. Is this the first time he’s done this? Or were there other spies before this? 2. Was he wittingly treasonous or an unwitting mark? 3. Who else assisted him and helped cover up that he was compromised? 4. Who else in the pro-China camp is compromised in this way?
So Jensen was having an affair with a customer, who also happened to be a Chinese spy, for months. Lied to the board. Lied to journalists. Meanwhile pushing to sell maximum of cutting edge AI chips to China against advice of ~entire natsec community. Can't make this shit up.
3
44
1,142
Oh.
It sounds like US officials have known about Jensen’s affair with a top customer since May, but the public (and Nvidia’s board!!!) only learned about it because of the excellent NYT reporting. Hugely impactful journalism.
2
21
942
It now appears that NYT is indicating the reason NVIDIA’s CEO was left off of President Trump’s China trip initially was because he is compromised by China, in the judgment of the National Security Advisor and Secretary of State.
NEW: Nat Sec Adviser Marco Rubio oversaw drafting of a highly classified memo this spring that the intel community assessed a woman who had established a close relationship w Jensen Huang was working for China's Ministry of State Security nytimes.com/2026/11/08/us/po…
2
15
210
16,064
I often disagree with Neil, so I want to flag here that I STRONGLY agree with him: we desperately need a definition of RSI that has gone through a bunch of rounds of lawyers kicking its tires. You're a law student, want to get cited for forever? Write this as your journal note.
Replying to @woodilicious
That’s not a definition that works in legislation.
4
10
3
175
7,541
It’s fascinating how many bots replied to this tweet. Wonder whose they are? (I’m not wondering.)
It is entirely possible (perhaps even likely!) that Jensen Huang is a victim, not a knowing agent run by this Chinese intelligence officer. But either way, he is compromised, and it is not tenable for him to remain in his role. ft.com/content/a55a2a22-4479…
1
1
28
1,150
It is entirely possible (perhaps even likely!) that Jensen Huang is a victim, not a knowing agent run by this Chinese intelligence officer. But either way, he is compromised, and it is not tenable for him to remain in his role. ft.com/content/a55a2a22-4479…
36
13
6
179
26,698
One thing that's fascinating is that if you point out to a McKinsey partner on LinkedIn that you think they shouldn't partner with OpenAI because OpenAI doesn't "adhere to the highest professional standards" (A core McKinsey value), they'll block you. Apparently, the obligation to dissent isn't something the Firm believes in any more. Kinda sad.
1
41
2,012
Many people feel like they can't "do AI policy" unless they live in DC or the Bay Area. Not so! You know what we need desperately right now? A bunch of people who care about AI policy in every state in the country, ESPECIALLY the small ones.
CALL YOUR CONGRESSMAN. Tell them to get ASI banned immediately. Don't let your families die because you were worried about the potential awkwardness of a phone call.
1
4
1
55
1,778
Since August, a team within METR has worked with internal Anthropic information to assess whether AI itself is accelerating AI development. We are now publishing a report summarizing the results of this risk assessment.
2
1
73
3,523
I criticized one METR chart design choice yesterday, so let me praise another today: burning in the “fake data” label in addition to writing “illustrative sketch, not real data”, is very good. Most people don’t read those sorts of labels unless you make it glaringly clear
Replying to @METR_Evals
2. We believe that a majority of this acceleration in capability improvement is likely now driven by AI being used in AI R&D. We aren’t sure of this, though, and did not find consensus about this among Anthropic’s staff. Non-AI factors likely played a large role in late 2025.
1
40
1,089
I think METR making this a log scale instead of a linear scale was a mistake.
the self-reported uplift from Model 4 is insane
8
6
1
222
11,654
How the hell did the fuzzers miss this?
The bug is like a joke. On Alpine Linux, type 257 to 1024 characters of anything and you get root. 💀
4
8
576
This is seriously good news.
Today, we released the National Policy for Nucleic Acid Synthesis Screening to protect American biotechnology innovation while reducing the risk of misuse. This policy establishes a new, government-wide screening requirements for synthetic DNA and the equipment used to make it. It will ensure companies selling nucleic acid sequences and synthesis equipment know their customers and prevent the malicious use of these materials and tools. This policy strengthens our Nation’s biosecurity posture by establishing rigorous U.S. government-wide procurement screening and physical security standards for synthetic biology. It replaces outdated “self-attestation” systems with robust, independent verification. It creates a dynamic, SI-powered screening layer that assesses the biological risk from nucleic acid sequences, allowing the U.S. to anticipate and prevent malicious use and help stop dangerous gain-of-function research. @NIST will issue rigorous new screening standards and @HHSGov and @USDA will certify synthetic biology providers. Together, these measures will strengthen American biosecurity, safeguard responsible innovation, and establish a global standard for safe and transparent synthetic biology tools and research.
1
18
611
One conversation I had yesterday with a colleague, who's never known anything other than the AI and tech industries: There genuinely are other industries that are better at this sort of thing! You know what Ford never does? Wake up the day after an election and go, "whoops, we never thought about that!" Their plans are measured in centuries!
NEW: The AI industry is scrambling to repackage a $10 trillion infrastructure buildout to appease a hostile electorate, that includes: -No more non disclosure agreements -Money for schools and cities -Environmental concessions -Labor protections -Codified obligations over voluntary frameworks -Job training programs The bottom line: The race is on to see whether Democrats can translate election momentum into binding restrictions or the industry tries to head off tougher rules. axios.com/2026/11/05/ai-data…
12
20
2
407
27,182
(Admittedly, Henry Ford Sr. probably only said, "Their plans are measured in centuries" while publishing his 91-week anti-Jewish conspiracy op-ed series in The Dearborn Independent, but you know, his successors have been rather less bigoted.)
2
46
2,557